In this thesis, we focused on the semantic interoperability problem for constrained devices in the Internet of Things(IoT). In IoT, there are enormous devices of differ- ent capabilities. A huge number of these devices are constrained devices and only produce data, then there is an intermediate layer of devices with sufficient capa- bilities to produce and consume data, and then on top are the servers with higher capabilities to facilitate these devices. Together they all create an Internet of Things ecosystem. In this thesis, we named the servers by their corresponding roles and used the term Endpoint for the all other types of devices. In the beginning, we discussed the technologies related to the Internet of Things to grasp a bigger pic- ture of the versatile nature of the IoT. The thesis introduces a new classification of IoT protocols and technologies to provide a better perspective from IoT point of view. This classification discuses everything from hardware communication to application-level interoperability. In order to focus on semantic interoperability, we divided the interoperability into three layers, which helped us to isolate semantic interoperability from it prerequisites layers of technical interoperability and syntac- tic interoperability. The thesis discussed the existing solutions to semantic inter- operability problem e.g. proxy gateways, unified data models & frameworks, and ontologies. All the above-mentioned methods have both positive aspects and few limitations. The main limitations of these methods are a single point of failure, scalability issues, integrating semantic interoperability with syntactic interoperabil- ity (that leads to the problem of full-scale adaptation of the proposed solution) and, resource-intensive solutions that are not feasible for constrained devices. Our con- clusion with the existing approaches is that they are applicable in IoT network with specific capabilities and requirements. The research goal of the thesis is to create a generic solution that could adapt to the dynamic behavior of the IoT ecosystem. Looking beyond the limited scope of the present IoT technologies, we found tech- nologies from the semantic web making their way to the Internet of Things, e.g. the Thing Description from the Web of Things project by W3C. Thing description is a great model for representing resources and interactions of any physical and virtual
devices. Using thing description to model our endpoints, we create a solution that can provide semantics for the data exchanged between endpoints. The core part of the solution is a server named "Interoperability Server" that provides semantic information about the data. Another server called "Resource Directory" provides the resource discovery functions for the IoT network. The chapter 4 describes the proposed solution in detail. However, the key aspects of this solution are,
• It does not limit the endpoints to use a specific data model or information model. An endpoint with any data model or information model can be repre- sented by Thing Description.
• It supports the hierarchical communication model where the endpoints at lower layers have fewer resources than upper layer endpoints. These resource- constrained endpoints can be managed by a third-party inside Interoperability Serve and Resource Directory to provide the semantics of their data for other endpoints.
• The registration with Interoperability Server and Resource directory can be Endpoint originated or managed by the third party as mentioned earlier. It decreases the overhead of registration for constrained devices.
• The endpoints can cache the translation lookup information for a specified period to reduce the lookup overhead and save power on Endpoint.
• The translations are provided on per resource basis for source and target End- point. This feature provides a security measure, by not exposing the whole Thing Description of the target Endpoint to the source Endpoint of the re- quest.
• The solution can withstand the inconsistent nature of IoT network, where the devices can join and leave the network at any time.
5.1
Discussion
The aim of the thesis was to create a semantic interoperability solution for IoT network. The thesis presented a solution to cater the said problem. The thesis has tried to answer the following main questions:
Question: What is a possible semantic interoperability solution that is independent of underlying technical and syntactic technologies of an IoT endpoint? Answer: The implemented solution describes every endpoint with Thing Description. The use of Thing Description provides the ability to represent any physical or virtual device of any capabilities independent of their underlying data models and infor- mation models. The working principles of this solution make it independent of any communication technology used to transfer data.
Question: How the presented solution provide semantics? Answer: The semantic interoperability means to agree on common meanings of data. To agree on common meanings of data in an IoT network, endpoints need the meta-data of the data. The solution presents a way to provide the meta-data of the output through translations. The presented solution is also applicable for cases when the endpoints e.g. sensors, actuators do not have enough resources to provide the meta-data information by themselves.
Question: How the presented solution meets the needs of resource-constrained IoT endpoints? Answer: The presented solution requires every endpoint must be reg- istered with Interoperability Server. For resource-constrained endpoints that do not have capabilities to register themselves, registration through third-party is sup- ported. Through this approach, all the resource constrained devices have repre- sentation in Interoperability Server, and the meta-data of their resources is readily available to other endpoints.
Question: Can the presented solution can be used in IoT Networks? Answer: Yes, the solution can be used to work in existing IoT networks facing problems of semantic interoperability. Any use case that involves different kinds of IoT endpoint or same kind of IoT endpoints with different standards (e.g multiple air pressure sensors from different vendors and possibly each vendor has its propriety standard) or a network where IoT endpoints leave and join very frequently, is a good candidate to deploy this solution. The data consumer endpoints of such networks (e.g a controller or a server using data from multiple sensors in an office) can have access to the meta-data of the requested data through Interoperability Server.
5.2
Proposed Future Work
We tried to design the proposed solution to be generic, simple, low network overhead, low power consumption and easy to adapt. To make this all possible, there are still improvements and future work required. Following are some key improvements and future work possibilities we propose.
The solution needs to implement a registration expiry or re-registration mechanism to remove dead endpoints from the Interoperability Server and Resource Directory. As mentioned in chapter 4, the solution implements the manual approach to create a translation. This approach was adopted considering the scope of the thesis and amount of work required. So, the main improvement and future work required is to implement an automatic translation creation mechanism. Upon our investigation, we have found that it can be made possible using the Ontologies. The Ontologies are already used in Semantic Web for semantic queries. Using the Ontologies with the machine learning and latest artificial intelligence technologies can produce the required functionality of automatic translation creation upon request.